Dofirmsrespondtogenderpaygaptransparency?...I Introduction Gender pay disparities characterize...
Transcript of Dofirmsrespondtogenderpaygaptransparency?...I Introduction Gender pay disparities characterize...
Do firms respond to gender pay gap transparency?∗
Morten Bennedsen† Elena Simintzi‡ Margarita Tsoutsoura§
Daniel Wolfenzon¶
November 5 2018
Abstract
We examine whether pay transparency closes the gender pay gap and affectsfirm outcomes. The paper exploits a 2006 legislation change in Denmark thatrequires firms to provide gender dis-aggregated wage statistics. Using detailedemployee-employer administrative data we find that the law has an effect inreducing the gender pay gap, primarily through slowing the wage growth formale employees. This effect is more pronounced for firms whose managers havemore daughters, presumably due to the effect of daughters on managerial pref-erences, and for industries with higher gender pay differentials pre-treatment.Such changes in firm wage policies following the passage of the law are asso-ciated with negative outcomes on overall firm productivity, but also with areduction in firm wage bill, resulting in no significant effects on firm profitabil-ity.
∗We thank Daniel Ferreira (discussant), Camille Hebert (discussant) and Rebecca Zarutskie (dis-cussant). We are grateful for excellent comments from seminar participants at UC San Diego, IESEBarcelona, Columbia Business School, Cornell University, Darden, INSEAD, UIC, LBS SummerFinance Symposium, Stanford SITE conference, CEPR Incentives, Management and Organization2018 conference, and the Corporate Finance Beyond Public Companies conference. We thank BrianJ. Lee for excellent research assistance, He Zhang for data management, and help with data under-standing from Mona Larsen (VIVE), Helle Holt (Vive) and Maria Boysen (Statistic Office Denmark).We are grateful for financial support from the Danish National Research Foundation (Niels BohrProfessorship).
†Niels Bohr Professor, University of Copenhagen and André and Rosalie Hoffmann Professor,INSEAD
‡Kenan-Flagler School of Business, University of North Carolina
§Cornell University and NBER
¶Columbia University and NBER
– 1 –
I Introduction
Gender pay disparities characterize labor markets in most developed countries
(OECD, 2015).1 When a man earns 100 dollars, a woman earns 77 in US (Goldin
(2014)), 78.5 dollars in Germany, 79 dollars in the UK and 83.8 on average across
EU countries (Eurostat, 2016). Recent proposals across many countries focus on pay
transparency to promote equal pay.2 However, evidence on the effect of transparency
on gender pay disparities on employee and firm outcomes is limited. In this paper,
we draw insights from a regulation change in Denmark to study how transparency
through gender based wage statistics may affect firm wage policy and outcomes.
There is an ongoing debate about disclosing gender wage gaps. Governments
often propose transparency as a tool to encourage firms to reduce the wage gap
between men and women. Unions and employee groups representing women also seem
to believe that secrecy on pay contributes significantly to unequal pay for women.3
Opponents of pay transparency argue that disclosing gender pay comes as a challenge
to firms as it lacks practical utility, increases administrative burden and violates
employee privacy and confidentiality.4 The effect of transparency on gender pay
disparities is an open empirical question, as it is unclear whether transparency will
incentivize firms to respond by adjusting their compensation policies and, if they do
1http://www.oecd.org/gender/data/genderwagegap.htm
2In the United Kingdom, employers of firms with more than 250 employees have to publish genderbased wage statistics from April 2018. In Germany, employees have the right to know median salaryfor a group of comparable employees in firms with more than 200 employees. An executive ordersigned by the U.S. government in 2016 required large companies to report salary data broken downby gender, starting in 2017 but the rule got halted by the succeeding administration.
3AFL-CIO runs a petition campaign as a response to the halt of the equal pay initiative thatwould have required large corporations to report pay data by gender to the Equal Employment Op-portunity Commission. https://actionnetwork.org/petitions/tell-the-eeoc-we-need-the-equal-pay-data-collection?source=website. The Institute for Women’s Policy Research in a survey documentsthat 60% of employees are discouraged or prohibited from sharing wage information and concludesthat pay secrecy is key to gender gap in earnings (IWPR, 2014).
4See for example, a letter representing employers against a bill in California that requires largefirms in the state to file reports detailing the gender pay gap for people working in the same position:http://blob.capitoltrack.com/17blobs/e3526ab2-1360-4461-a1d3-b0580abe6172
– 2 –
so, by which margins.
Studying this question empirically requires addressing two key challenges: finding
exogenous variation in transparency at the firm level as well access to wage data
at the individual level. To address these difficulties, we exploit a 2006 legislation
change in Denmark that requires firms with more than 35 employees to report salary
data broken down by gender for employee groups large enough so that anonymity
of individuals can be protected. Under the 2006 law, firms have the duty to inform
their employees of wage gaps between men and women and explain the design of the
statistics and the wage concept used. In addition, we use data at the employee-firm
level from the Danish Statistics matched employee-employer administrative dataset.
We employ a difference-in-differences approach, where treated firms employ 35-50
employees on average from 2003 to 2005, the years prior to the introduction of the
law, and control firms employ 20-34 workers on average. Using detailed worker-firm
data, we then compare the change in employee and firm outcomes around the passage
of the law for treated firms relative to control firms.
Our sample firms pay their male employees a 18.9% wage premium before the
regulation is introduced. This gender pay differential is not driven by differences in
demographics, work experience, macro trends or selection into specific occupations as
we are able to absorb such variation using detailed controls in our specifications. We
find that transparency lowers the gender wage gap: it is reduced by 1.4 percentage
points in treated relative to control firms, or a 7% reduction relative to the pre-
regulation wage gap. Uncovering the source of adjustment, we show that wages for
all employees (both male and female) increase over time; however, male wages in
treated firms increase by less and female wages in treated firms (weakly) increase by
more, thus contributing to an overall decline in the male wage premium for treated
firms following the law change.
To further account for important drivers of gender pay disparities (Blau and
Kahn, 2017), not necessarily related to pay discrimination, such as differences in
– 3 –
skills, selection of women in certain occupations, industries or firms, we employ the
full capacity of the firm-worker administrative data and include interacted firm and
individual fixed effects in our specifications, in addition to individual time-varying
controls and year fixed effects. We find that wages of male employees in treated firms
are lower by 1.7 percentage points relative to male employees in control firms and
this reduction in wages is statistically significant. On the contrary, we find a positive
but not significant relation for female workers. Overall, we show that men’s wages
grow by 2 percentage points less relative to women in treated as opposed to control
firms following the law.
We provide additional analysis that further supports a causal interpretation.
First, we estimate the effect of the law by-year and find no evidence of pre-treatment
trends. Second, we perform placebo tests using alternative employee size cutoffs to
define treatment and find no significant effects. Third, we show our results are robust
to estimating our analysis within firm-years, by including interacted firm and year
fixed effects. As such, we absorb any time-varying shocks at the firm level that may
be correlated with changes in firm labor demand, further alleviating concerns that
time-varying differences between treated and control firms are driving our findings.
Fourth, we get similar results when we use hourly wages as our compensation mea-
sure or when we also consider non-salary compensation components, such as bonus
compensation.
Besides documenting that pay transparency against gender pay discrimination is
effective in changing compensation within firms, we provide evidence on how the law
affected employee reallocation. Using the same empirical design at the firm level, we
show that treated firms hire more female employees as compared to control firms in
specifications that include firm and year fixed effects. This is in line with an argu-
ment that the supply pool of female employees increases as gender pay transparency
improves and thereby gender pay gap closes. On the contrary, we do not find a sta-
tistically significant effect on female employees’ departures following the law passage.
We also find that the law has spillover effects on promotion decisions to the favor
– 4 –
of female employees. We find that women tend to be promoted from the bottom of
the hierarchy to more senior positions, while we do not find any significant change in
promotions for male employees.
In additional tests, we examine the implications of gender pay transparency on
firm outcomes. For one, our findings suggest that the law resulted in lower wage
growth for treated firms as firms pay relatively lower wages to male employees. In-
deed, we confirm our finding at the firm level where we show a negative and significant
effect on treated firms’ wage bills, which are lower by 2.8%, as compared to control
firms. Moreover, the law significantly affects employee productivity. A priori, the
effect on productivity is ambiguous. If information on gender pay gaps will lower
job satisfaction for those employees paid below their reference group—either because
female employees learn of the pay gaps, or because male employees are dissatisfied
with firms giving them lower pay increases as a response to the law— then we should
expect to see a negative effect on firm productivity (Akerlof and Yellen, 1990). If
instead, the reduction in wage disparities will create a sentiment of fairness among
workers, employee productivity may increase. We present evidence suggesting that
productivity (measured as the logarithm of sales over employees) drops by 2.5% rel-
ative to controls following the passage of the law. As such, the negative effect on
productivity is offset by firms’ lower wage costs, resulting in no significantly different
effects on firm profitability.
Finally, we explore potential mechanisms explaining firm responses to the law. We
argue that managerial preferences and pre-law industry gender pay differentials are
non-mutually exclusive factors that can help explain the way firms adjust to increased
transparency following the regulation. Fist, consistent with the idea that managerial
styles affect corporate policies (Bertrand and Schoar, 2003), we argue that manage-
rial preferences may affect how managers respond to the regulation for transparency
on pay equity in the workplace. We use the finding in the literature that men par-
enting daughters are more likely to adopt pro-women preferences because they have
experienced a higher degree of "female socialization" (Dahl, Dezső, and Ross, 2012).
– 5 –
Indeed, we show that managers with more daughters exhibit higher sensitivity to the
law passage closing the gender pay gap by increasing female compensation relatively
more.
Second, firms’ responses may depend on the pre-law gender pay inequality if
increased transparency led to greater accountability. Using the pre-law within occu-
pation gender pay inequality in the industry, as a proxy for pre-law within occupation
firm level wage differentials, we see a stronger adjustment if within occupation in-
equality at the industry is higher prior to the law introduction.
The paper contributes to the literature on the effects of mandated pay trans-
parency. Using government employees in California Card et al. (2012) shows that
after government employee salaries are published online, aggregate worker satisfac-
tion drops. Mas (2017) shows that top earners in municipal jobs experience a drop
in wages following the public disclosure of wages, which he argues is primarily due
to public aversion to visibly exorbitant salaries. Both of these studies focus on the
public sector sector.5 We provide the first evidence based on private firms.
Moreover, the paper relates to a growing literature studying how wage disparities
within the firm may affect employee or firm outcomes. For example, Mueller et al.
(2017) find that firms with higher pay inequality exhibit larger equity returns sug-
gesting that differences in pay inequality across firms are a reflection of differences
in managerial talent. Breza et al. (2018) uses a sample of workers in an Indian man-
ufacturing plant to show that information on how much peers are earning, relative
to one’s own salary, might generate negative feelings and reduce job satisfaction.
However, these papers do not link wage transparency to firm outcomes. In addition,
our study is the first study with a specific focus on gender disparities—an issue of
debate.
5The mechanisms that affect wage setting in the public sector might be different than thoseaffecting wages in the private sector, for example in the public sector public pressure and publicaversion to high compensation or inequalities might play a larger role than in the private sector.
– 6 –
The paper also relates to the vast literature that studies the sources of gender
pay disparities.6 Although most work in the area has focused on determinants of the
gender pay gap, there is limited evidence on what might be the possible solutions.
After accounting for the common drivers of gender pay gap in the literature, we
provide support for pay transparency as a potential avenue to mitigate gender pay
gaps within organizations.
Our paper contributes to a growing literature on gender and organizations that
point to biases facing women in the professional workforce. Egan, Matvos, and Seru
(2017) show that female advisers face harsher outcomes following misconduct, but
this effect is mitigated for firms with more female executives. Adams and Ragunathan
(2017) show that gender barriers tend to discourage women from working in finance.
Duchin, Shamir, Patriat, Kim, Vitek, Sapiro, and Harel (2018) show that female
division managers are allocated less capital, especially in firms where CEOs grew up in
male-dominated families. Tate and Yang (2015) show that male leadership cultivates
a less female-friendly culture within firms. Our findings suggest that regulatory
mandates on pay transparency, as a means to overcome biases against women in the
workforce, may be effective in closing the gender pay gap.
II The Law
On December 7, 2005, the Minister for Employment submitted a proposal to
Parliament to amend the Equal Pay Act. The proposal was adopted by Act no. 562
on June 2006. The goal of the law was “to promote visibility and information about
wage differentials.” The law stated that an employer with a minimum of 35 employees
and at least 10 employees of each gender within a 6-digit DISCO code (occupation
classification code) shall each year prepare gender-segregated wage statistics for the
purpose of consulting and informing the employees of wage gaps between men and
6See, for example, Goldin (2014) and Blau and Kahn (2017) for a review of the literature.
– 7 –
women in the firm.7 The statistics had to be made available to the employees through
the employee representatives; they did not need to be made available to the general
public. The law also offered an alternative choice to employers by permitting to
replace gender based wage statistics with an internal report on equal pay. This report
had to include a description of the conditions that are important for determining the
wage and establish an action plan for equal pay to be implemented within a three-
year period. The law establishes that non-compliance by firms would be punished by
a fine. The new provisions came into force on January 2007.
III Empirical Design
To estimate the effect of gender pay transparency on employee pay and other firm
outcomes, we employ a difference-in-differences approach. Our treated firms are firms
that employ 35-50 employees the year prior to the introduction of the law and the
control firms are those that employ 20-34 workers. We are taking a narrow window
around the 35-employee cutoff so that we compare firms of similar size.
We design our empirical strategy around the 35 threshold and do not take into
account the criterion that firms have at least ten male and ten female employees in
one six-digit DISCO code. The reason is that firms do not typically have DISCO
code information. According to the Danish Employer Association "some firms may
still internalize the law even when do not satisfy the second criterion (DISCO)."8. In
fact, 35% of firms that reported gender disaggregated wage statistics with the DA
did not satisfy the second criterion; yet all of them had more than 35 employees.
In addition, this is consistent with how the law was interpreted more widely. The
description of the law by the EU and the ILO only mentions the requirement of the
7The requirement does not extend to companies in the fields of farming, gardening, forestry andfisheries.
8Private communication
– 8 –
threshold of 35 employees.9
We use a panel of employee-firm-years to test whether disclosure of information on
wages by gender has real effects on firms’ compensation policies. We, thus, compare
firms with employees just above the employee threshold defined by the law with those
just below the threshold. We estimate OLS regressions of the following form:
log(wage)ijt = αij + αt + γ1Xit + γ2Xijt + β1I(Treatedijt) + β2I(Postijt)
+ β3I(Maleijt) + β4I(Treatedijt × Postijt)
+ β5I(Treatedijt ×Maleijt) + β6I(Postijt ×Maleijt)
+ δ I(Treatedijt × Postijt ×Maleijt) + εijt (1)
where i, j, and t index firms, individuals and years; post takes a value of 1 for
2006, 2007, and 2008 and a value of 0 for years 2003, 2004 and 2005; Xit and Xijt
capture time-varying firm- and individual-level control variables, respectively. We
add controls for time-varying individual characteristics (age, experience, education),
following Blau and Kahn (2017). αt is year fixed effects to absorb aggregate macroe-
conomic shocks. We also include interacted individual firm fixed effects, αij. The
individual times firm fixed effects allows us to control for time-invariant person char-
acteristics, time invariant firm characteristics as well as the match between firms and
workers. Effectively we are able to identify the effect on wages, clean of composi-
tional changes at the firm. Our empirical strategy allows to account for important
drivers of gender pay disparities (Blau and Kahn, 2017), not necessarily related to
pay discrimination, such as differences in skills, selection of women in certain occu-
pations, industries or firms. Standard errors are clustered at the firm level. We start
9ILO describes the law as: “Employers employing 35 or more workers are required to pre-pare annually gender-disaggregated statistics or, alternatively, an equal pay report and actionplan.”European commission directorate for internal policies issued a report on policies on Gen-der Equality in Denmark describing the law: “Since 2007, companies with 35 employees or moreshould carry out gender disaggregated pay statistics and elaborate status reports on the efforts topromote equal pay in the workplace.” (European Commission, 2015).
– 9 –
our sample in 2003 to provide sufficient years to estimate the baseline effect for each
firm-employee and end in 2008 to avoid overlap of our sample with the financial crisis
which had economy-wide effects on wages.
We also examine the effect of the law on firm outcomes, such as hiring decisions,
productivity and profitability. Using instead a panel of firm-years, we estimate OLS
regressions of the following form:
Yit = αi + αt + γXit + β1I(Treatedit) + β2I(Postit)
+ δ I(Treatedit × Postit) + εit (2)
where i, and t index firms and years; post takes a value of 1 for 2006, 2007, and
2008 and a value of 0 for years 2003, 2004 and 2005; Xit captures time-varying firm-
level control variables. The coefficient of interest δ captures the differential effect of
the law on the dependent variables for treated and control firms. αt is year fixed
effects to absorb aggregate macroeconomic shocks. We also include firm fixed effects,
αi to control for time-invariant firm characteristics. Standard errors are clustered at
the firm level.
IV Data and Sample Description
IV.1 Data sources
Our main dataset is the matched employer-employee dataset from the Integrated
Database for Labour Market Research (IDA database) at Statistics Denmark. In
addition to the employer’s identification number (CVR), and employee identification
number (CPR), the IDA dataset contains detailed information for employees’ com-
pensation, demographics and occupation. For compensation we have information on
employees’ wage and bonus. Furthermore, for each employee we observe their age,
gender and education as well as their position in the organization.
– 10 –
This information is combined with firm-level outcomes from the Danish Business
Register. This dataset covers all firms incorporated in Denmark and includes the
information these firms are required to file with the Ministry of Economics and Busi-
ness Affairs, including the value of total assets, number of employees and revenues.
Even though most firms in this dataset are privately held, external accountants audit
firm financial information in compliance with Danish corporate law. We link infor-
mation in the firm-level dataset to the the matched employer-employee dataset using
the firm identifier (CVR number).
IV.2 Sample construction and summary statistics
We start with the universe of public and private limited liability firms in Denmark
and their employees included in the Integrated Database for Labour Market Research.
For ease of comparison, for the employee level outcomes we focus on full-time workers,
excluding CEOs and board directors. We drop firms in industries unaffected by the
policy (farming, gardening, forestry and fisheries). We require firms to have financial
information which results in dropping 0.8% of firm years over our sample period.
Table 1 presents summary statistics for the treated and control firms in our sample
over the 2003-2005 period prior to the law passage. Panel A presents employee level
characteristics and Panel B presents firm level characteristics. The average annual
(hourly) wage for employees in the treated firms is $55 thousand ($34.4), while for
the control group is $53 thousand ($33.5). The average employee in the sample is
40 years old and has 17 years of work experience in both treated and control groups.
On average, 25% of employees in treated or control group hold a college degree.
Treated firms are larger than control firms by definition. For example, the average
treated firm has 42 employees pre-treatment, assets of $7.2 million and sales of $11.68
million as compared to 26 employees, $6.1 million in assets and $7.73 million in
sales for control firms. However, firms are similar in terms of their pre-treatment
productivity, cost structures, and the gender composition of their employees with
– 11 –
70% male employees on average.
Panel B2 shows the pre-law wage premiums10 for treated and control firms. We
estimate the male wage premium both without controls and with adding year fixed
effects, time-varying individual characteristics (age, experience, education), control
for sales, occupation fixed effects and industry fixed effects. In our stricter specifi-
cation we observe that the male wage premium is 20.2% for the treated group and
17.9% for the control group prior to the regulation.
V Results
V.1 Wages
Our goal is to identify the effect of transparency on firm compensation policies
and the relative pay of men and women. Before we present our OLS results, we show
univariate tests that demonstrate the main effect. Table 2 presents the average log
wage in years 2006-2008 minus the average log wage in 2003-2005, the three years
prior to the passage of the law. Wages increase for all employees, irrespective of their
gender, in both the treated and the control group. However, male employee wages
grow by less in treated firms as compared to control firms and the difference in the
average increase between the treated and the control groups (-0.0156) is statistically
significant at the 1% level. Similarly, both treated and control firms’ female employees
receive higher wages on average following the reform. The increase is higher, however,
for the female employees in the treated firms as compared to control firms, although
this difference is not statistically significant. These univariate comparisons suggest
that the reform results in lower wage growth for male employees and higher (although
not statistically significant) wage growth for female employees. This results in about
2 percentage points lower wage increase for male versus female employees in treated
10The male wage premium is the estimate of the coefficient on a male dummy in a log wageregression
– 12 –
relative to control firms.
We next turn to our regression analysis and use detailed employee-firm wage data
to account for the possibility that compositional changes at the firm level may affect
the observed differences in wages. We thus estimate the effect of disclosing gender
pay disparities on wages of a given individual within a treated firm as compared to
an individual in a control firm. Table 3 reports the results. In our regressions, we
include firm-individual fixed effects to control for firm and individual time-invariant
characteristics and the match between firms and employees, and year fixed effects to
absorb macroeconomic shocks.
Column 1 compares the effect of the law on wages of male employees in treated
firms as compared to male employees in control firms. Column 2 repeats this analysis
comparing instead wages for female employees. We find that wages of male employees
in treated firms grow relatively less as compared to male employees in control firms
following the passage of the law, similar to our univariate results in Table 2. The
effect is statistically significant at the 1% level and economically important. Treated
male wages are reduced by 1.67 percentage points more relative to male wages in
the control group. On the contrary, we find a positive but not significant coefficient
on treated firms’ female wages relative to control firms in column 2. In a triple
differences estimation in column 3, we compare the effect of the law on wages of male
employees as compared to wages of female employees following the passage of the law
relatively to a group of control firms. The triple difference coefficient shows that male
wages grow by 2 percentage points less then female wages in treated versus control
firms and the effect is statistically significant at the 1% level.
In columns 4-6, we repeat our estimation additionally controlling for firm size
(proxied by logarithm of sales) to control for the well documented employer size-
wage effect (e.g. Brown and Medoff (1989); Idson and Oi (1999)). Note the employer
size-wage effect would predict a positive effect for wages of larger firms (indeed the
coefficient for sales is positively correlated with wages and significant at the 1%
– 13 –
level)—the opposite from our findings that the law negatively affects wage growth
for larger firms. Including firm size is also important in our setting given the treated
group includes by construction larger firms. The estimated coefficients remain vir-
tually unchanged after controlling for firm size. In Internet Appendix Table IA1, we
repeat our analysis including in the sample only individuals that worked for the firm
at least one year in the pre-law period and one year in the post period. This test
further addresses concerns that changes in the composition of a firm’s employees are
driving our results. The estimated coefficients are qualitatively similar, except the
positive effect on female wages is now weakly statistically significant.
V.2 Identification concerns
Our analysis allows us to absorb a lot of unobserved variation by including de-
tailed individual controls and interacted person and firm fixed effects. Moreover,
the fact that our estimated effect on wages is concentrated on male employees (as
opposed to all employees) mitigates identification concerns related to omitted vari-
ables which might correlate with firm size, e.g. firms of different sizes adjusting their
compensation differently to their investment opportunity sets. To drive our findings,
an omitted variable would not only need to be correlated with wages, but also differ-
entially affect male wages across different firms. In this section, we provide further
evidence that our results are consistent with a causal interpretation.
Table 4 shows year-by-year coefficients for male (column 1) and female (column
2) employees before and after the passage of the law. We find no significant difference
in the evolution of wages at treated and control groups prior to the adoption of the
law. Column 3 presents year-by-year estimates of the triple interaction coefficients
and also shows that male wage growth is significantly lower in 2007 and 2008 by
2.1 percentage points and 1.9 percentage points, respectively, as compared to female
wage growth in treated versus control firms, while there is no significant difference
pre-treatment.
– 14 –
To further alleviate the concern that an omitted variable differentially lowers male
wages at larger firms, we create placebo tests where we use alternative employee size
cutoffs to define treatment. In columns 1-3, Table 5, we define placebo treated firms
to be firms with 20-35 employees in the 2003-2005 period and placebo control firms
those firms with 5-19 employees. In column 4-6, we use 50 employees as the cutoff
and thus, placebo treated firms are those firms with 50-65 employees pre-treatment
and placebo control firms are firms with 35-49 employees pre-treatment. In columns
7-9, we instead use a cutoff of 65 employees and thus placebo treated firms are those
firms with 65-80 employees pre-treatment and placebo control firms are firms with
50-64 employees. We are unable to replicate our baseline findings when considering
these alternative cutoffs, consistent with the fact that the effect is unique to the 35
employee cutoff as described by the law.
Moreover, we repeat our baseline analysis additionally controlling for interacted
firm and year fixed effects. These controls allow us to absorb any time-varying
changes at the firm level that could be driving our results. To include firm-year
fixed effects, we need variation within firm-year and as such, we can only repeat
specifications similar to those in column 3, Table 3, where we provide a triple differ-
ence estimate comparing the effect of the law between male and female employees in
treated versus control firms. Table IA2, in the Internet Appendix, repeats estimates
in Tables 3 and 4 and returns very similar estimates when we additionally control
for firm level shocks.
One potential concern with the interpretation of our findings could be that the law
resulted in men working less hours in treated firms as this could mechanically improve
the gender wage statistics. To examine whether this alternative interpretation might
be true, we replicate Table 3 using instead employee hourly wages. In Internet
Appendix IA3, we show that the results are similar both in terms of economic and
statistical significance. The measure of hourly wages comes from a mandated pension
scheme introduced in 1964 - Arbejdsmarkedets Tillaegspension (ATP)- that require
all employers to contribute on behalf of their employees based on individual hours
– 15 –
worked. One caveat, however, as explained in Kleven, Landais, and Søgaard (2018),
is that this ATP based measure of hourly wages is based in bracketed hours worked
and it is capped, which is not the case for our baseline wages measure. Moreover,
we examine the possibility that the reduction in salary may be (partially) offset by
relatively higher bonuses offered to male workers in treated firms. Thus, in Internet
Appendix Table IA4 we estimate the effect of the law on employee total compensation
(wage and bonus payment). Bonus payments do not seem to materially affect our
estimates as shown by the coefficients in Internet Appendix Table IA4.
VI Pay by Hierarchy, Hiring and Promotions
Does increased transparency affect all employees in the firm, or do we observe
asymmetric responses by firms depending on employee hierarchy? In Table 6, we
examine the effect of the law on pay for managerial employees, at the top of the
hierarchy, and for employees in non-managerial positions at lower hierarchy levels.
IDA database provides information on the primary working position of the employee
and whether the employee is high-level employee, intermediate-level employee or low-
level employee. Columns 1-3 show the results for managerial employees in the high
hierarchy levels; columns 4-6 instead present results for non-managerial employees in
lower hierarchy levels. It can be observed that there is no significant relationship on
pay for either men or women at the top, while there is a strong and significant effect
for all other employees.11 These results are consistent with the fact that the law is
more likely to apply to employees compensated based on wages and not performance
pay.
Although the law seems to have a clear effect on wages, as intended by the reg-
11In unreported results we replicate this analysis defining firm hierarchies based on worker’s oc-cupations following Caliendo, Monte, and Rossi-Hansberg (2015) and Friedrich (2015). Hierarchicallayers group occupations in a systematic way in four layers to focus on vertical relationships betweentop managers, middle managers, supervisors, and workers. Using this alternative measure of firmhierarchies, we find consistent results that significance comes from the non-managerial layers.
– 16 –
ulator, this might not be the only response by firms. Changes in the way similar
employees of different gender are compensated might affect the demand or supply for
those employees, resulting in differences in hiring or departure rates. Alternatively,
the law mandate for fairer practices may have spillover effects on other firm decisions
such as employee promotions. We examine the effect of the law passage on each of
these different outcomes next.
We start by computing hiring rates for female employees at the three hierarchy
levels within the firm, as in Table 6. Thus, Joining rate is the total number of female
employees joining the firm-hierarchy in a given year t normalized by the total number
of employees joining. (By construction, hiring rates for men and women sum up to one
and thus, we only present hiring rates for female employees). We thus compare hiring
rates for women in treated versus control firms in a given hierarchy level following
the policy change, in a specification with firm and year fixed effects. We present
results in Panel A, Table 7. Conditional on hiring, we find that treated firms hire
a higher share of women in the intermediate hierarchy levels, which is statistically
significant at the 5% level. We also find an economically large effect for low hierarchy
levels although this effect is statistically noisier, and we observe no difference at the
top. The pre-law average female joining rate is 37% and 43.6% for intermediate
and low hierarchy levels respectively, and the joining rate of women increased by 4.4
percentage points and 2.5 percentage points, respectively. This finding is consistent
with the fact that firms are able to attract more female employees in positions where
they offer a fairer compensation.
Similarly, we define departure rates as the number of female employees leaving
in given firm-hierarchy-year normalized by the total number of employees leaving in
that firm-hierarchy-year. Our goal is to capture voluntary departures from the firm
rather than firings. Therefore from our measure we exclude departures where the
employee remained unemployed for more than a year. In Panel B, Table 7, we find
no statistically significant change in departure rates of males or females across firm
hierarchies. Interestingly, however, the departure rate for high-level female employees
– 17 –
is economically large. Although weak, this evidence suggests that women are more
likely to leave from positions where there was no adjustment in pay towards closing
the gender pay gap. Overall, these results suggest that women participation rates
increase in those occupations where male wage premium is reduced.
To examine firm promotion decisions, we define a dummy that takes a value of 1
if a given individual is promoted to a higher hierarchical level. The measure is thus
meaningful for the intermediate and low level employees. Table 8 present the results.
Columns 1-3 show that for intermediate level employees there is no change in their
propensity to get promoted to the highest hierarchy level after the passage of the
law. Columns 4-6 show instead that low-level female employees are more likely to
be promoted to higher hierarchy levels in treated firms after the passage of the law,
as compared to controls. The promotion probability before the reform is 2.2% for
males and 2% for females, and although it does not change for males, it increases by
1.2 percentage points for female employees. These results complement our previous
findings indicating that the law did not only have the intended consequences of
“fixing” gender pay disparities within the firm, but also improved female employees’
ability to climb up the corporate ladder.
VII Firm performance
We next explore whether the effect of the law on gender pay and employee reallo-
cation also affects firm productivity and profits. In doing so, we need to caution that
the average treatment effect we are able to estimate may be driven by both compen-
sation or compositional changes at the firm level as a result of the law, limiting our
ability to pin down the precise mechanism or responses by different employee groups
within the firm.
We perform our analysis at the firm level in a specification with firm and year
fixed effects. We report the results in Table 9, with and without controlling for firm
size. In columns 1-2, we examine the effect of the law on firm productivity of treated
– 18 –
firms as compared to the group of controls. The effect on productivity is theoretically
ambiguous. If information on gender pay gap will lower job satisfaction of female
employees, that should negatively impact their productivity. A similar effect should
be observed if male employees get dissatisfied with firms’ lowering their wages relative
to their peers. However, if increased transparency and firms’ responses create a senti-
ment of fairness among employees, then productivity should be positively impacted.
We observe that, on average, productivity (measured as the log transformed sales
per employee) drops by 2.5% in treated firms following the regulation, as compared
to controls, and this reduction is statistically significant at the 5% level.
However, this drop in productivity seems to be exactly offset by the lower wages
due to the fact that firms respond to the law by lowering male employee wages.
Indeed, columns 3 and 4, Table 9 show that the average wage per employee (log-
transformed) is reduced by 2.8%, canceling out the negative productivity effect. Note
we only observe a negative and significant effect on employee wages and not on other
labor costs, such as pensions and other social security costs, as the latter are not
directly impacted by the regulation.
These results can explain why we do not observe any significant effect on firm
profitability, in columns 7-8, Table 9. Because of the accounting identity, the effect on
profits must reflect some combination of the decrease in costs but also of the decrease
in revenues. Using profit per employee, as our measure of profitability, we find no
effect of the law on firm profits. A null result on profitability renders no support for
critics of the law who argue that disclosing pay gaps may be particularly costly to
firm profits.
VII.1 Mechanisms
We next explore potential channels explaining firm’s response to the mandate for
transparency following the passage of the law. We propose two non-mutually exclu-
sive mechanisms: 1) managerial preferences; and 2) pre-law gender pay differentials.
– 19 –
First, we propose that managerial styles may affect the way firms respond to
the law, as they have been shown to affect corporate policies (Bertrand and Schoar,
2003). To create a proxy for managerial preferences that would favor women, we
start from the finding in the literature that men parenting daughters are more likely
to adopt pro-women preferences (Warner, 1991; Warner and Steel, 1999; Oswald and
Powdthavee, 2010; Washington, 2008; Glynn and Sen, 2015; Cronqvist and Yu, 2017;
Dahl, Dezső, and Ross, 2012).12 As such, managers with daughters should not only
be more likely to follow fairer pay practices towards women,13 but they should also
exhibit greater sensitivity to the law passage.
To test this, we consider a firm’s managerial team to be the top five earners in
the firm in the 2003-2005 period. We exclude female managers and consider (up
to) five male managers.14 We then define a variable to be 1 if a male manager
has more daughters than sons, 0.5 if they have as many daughters as sons, and 0
otherwise. We average the above categorical variable for each firm’s managerial team
and define Female Child to be one if the firm average is above the sample median,
0 otherwise. We augment our baseline specifications by interacting Treated × Post
with Female Child. Table 10 presents the results. We find that firms where male
managers have more daughters pay higher wages to their female employees following
the law passage relative to controls, while we observe no significant difference for male
employees. In column 3, Table 10, we show that this results in closing the gender
pay gap by 2.4% more for treated firms whose managers tend to be more favorable
12Examples that support the female socialization hypothesis abound in the social sciences lit-erature. Washington (2008) and Glynn and Sen (2015) find that having a daughter increases thepropensity to vote liberally for members of the U.S. Congress or federal judges, respectively. Oswaldand Powdthavee (2010) show, more generally, that parents with daughters tend to be politicallymore left-oriented. Cronqvist and Yu (2017) show that CEOs with daughters are more likely tomake corporate social responsible decisions, especially related to issues concerning diversity, theenvironment, and employee relations.
13In unreported regressions, we confirm that managers that have more daughters tend to offerhigher wages to women in the pre-treatment period, but not to men, and this difference is statisticallysignificant.
1420% of top managers in our sample are women.
– 20 –
to women, as compared to controls, and this effect is statistically significant at the
5% level. In Internet Appendix Table ??, we instead construct the Female Child
measure based on the first-born child of the firm’s top managers, which is arguably
a more exogenous child gender measure (Bennedsen, Nielsen, Perez-Gonzalez, and
Wolfenzon, 2007; Cronqvist and Yu, 2017). The estimated coefficients are similar to
those reported in Table 10.
Second, we consider the role of the pre-existing gender pay inequality.Increased
transparency might lead to increased accountability thus firms with higher pre-
existing inequality might be more responsive to reduce pay inequality. Ideally, we
would would like to have a measure of gender pay inequality at the firm level but this
measure would be very noisy given that firms are unlikely to have a large number of
men and women in the same occupation so that we can have a meaningful measure.
Therefore we use the pre-law within occupation gender pay inequality in the industry,
as a proxy for pre-law within occupation firm level wage differentials
To this end, we define gender pay gaps at the industry-occupation level by com-
puting the median log difference in wages by gender at the industry-occupation-year
level and averaging over the pre-treatment period (2003-2005). We augment our
baseline specification by interacting Treated×Post with Ind. Gender Gap, the pre-
treatment industry-occupation gender pay differential. In Table 11, we show that
treated firms in industries with high gender disparities pre-treatment pay relatively
lower wages to male employees, although this difference is not statistically significant.
In contrast, they pay female employees more relative to controls, and this difference
is both statistically and economically significant. A one standard deviation increase
in pre-treatment industry gender pay gaps is associated with an increase in female
wages by 1.25%. Most importantly, in column 3, we show that gender pay gaps close
by more when pre-treatment inequality is higher. Specifically, a one standard devia-
tion increase in Ind. Gender Gap is associated with a 1.6% reduction in the gender
pay gap.
– 21 –
In sum, these results suggest different mechanisms at play that can plausibly
explain the observed response by firms to the increased transparency on gender pay.
A caveat of this analysis, however, is that these tests are suggestive and we cannot
assess the relative importance of each of these mechanisms.
VIII Conclusion
Reducing the gender pay gap has been at the epicentre of a heated debate among
academics and policy makers. Recently, governments around the world proposed
transparency as a tool to inspire firms to reduce the wage gap between men and
women. Nevertheless, there is no systematic study that examines the effects of in-
creased transparency of within firm gender pay disparities on firm wage policy and
outcomes.
Investigating empirically the effect of pay transparency rules as a measure to
reduce pay discrimination within firms is challenging as it requires exogenous find-
ing variation in transparency rules but also having detailed information of employee
wages. We overcome these hurdles by exploiting a 2006 regulation in Denmark that
requires certain companies to report gender-segregated wage statistics. Using de-
tailed employee-firm matched administrative data and using a difference-in-difference
methodology we find changes in compensation within firms. Specifically male em-
ployees experience slower wage growth relative to female employees. We argue that
firm, industry and managerial characteristics play a non-mutually exclusive role in
explaining firms’ response to the increased transparency.
– 22 –
References
Adams, R. B., Ragunathan, V., 2017. Lehman sisters. Available at SSRN .
Akerlof, G. A., Yellen, J. L., 1990. The fair wage-effort hypothesis and unemployment.
The Quarterly Journal of Economics 105, 255–283.
Bennedsen, M., Nielsen, K. M., Perez-Gonzalez, F., Wolfenzon, D., 2007. Inside the
Family Firm: The Role of Families in Succession Decisions and Performance. The
Quarterly Journal of Economics 122, 647–691.
Bertrand, M., Schoar, A., 2003. Managing with style: the effect of managers on firm
policies. Quarterly Journal of Economics 118, 1169–1208.
Blau, F. D., Kahn, L. M., 2017. The gender wage gap: Extent, trends, and explana-
tions. Journal of Economic Literature 55, 789–865.
Breza, E., Kaur, S., Shamdasani, Y., 2018. The Morale Effects of Pay Inequality. The
Quarterly Journal of Economics 133, 611–663.
Brown, C., Medoff, J., 1989. The employer size-wage effect. Journal of political Econ-
omy 97, 1027–1059.
Caliendo, L., Monte, F., Rossi-Hansberg, E., 2015. The anatomy of French production
hierarchies. Journal of Political Economy 123, 809–852.
Cronqvist, H., Yu, F., 2017. Shaped by their daughters: Executives, female social-
ization, and corporate social responsibility. Journal of Financial Economics 126,
543–562.
Dahl, M. S., Dezső, C. L., Ross, D. G., 2012. Fatherhood and Managerial Style: How
a Male CEOâĂŹs Children Affect the Wages of His Employees. Administrative
Science Quarterly 57, 669–693.
– 23 –
Duchin, Y., Shamir, R. R., Patriat, R., Kim, J., Vitek, J. L., Sapiro, G., Harel,
N., 2018. Patient-specific anatomical model for deep brain stimulation based on 7
Tesla MRI. PLOS ONE 13, 1–23.
Egan, M. L., Matvos, G., Seru, A., 2017. When Harry Fired Sally: The Double
Standard in Punishing Misconduct. Working Paper 23242, National Bureau of
Economic Research.
Friedrich, B., 2015. Trade Shocks, Firm hierarchies and Wage inequality. Economics
working papers, Department of Economics and Business Economics, Aarhus Uni-
versity.
Glynn, A., Sen, M., 2015. Identifying Judicial Empathy: Does Having Daughters
Cause Judges to Rule for Women’s Issues? American Journal of Political Science
59, 37–54.
Goldin, C., 2014. A grand gender convergence: Its last chapter. American Economic
Review 104, 1091–1119.
Idson, T. L., Oi, W. Y., 1999. Workers are more productive in large firms. American
Economic Review 89, 104–108.
Kleven, H., Landais, C., Søgaard, J. E., 2018. Children and gender inequality: Ev-
idence from Denmark. Working Paper 24219, National Bureau of Economic Re-
search.
Mas, A., Moretti, E., 2009. Peers at work. American Economic Review 99, 112–45.
Mueller, H. M., Ouimet, P. P., Simintzi, E., 2017. Within-firm pay inequality. Review
of Financial Studies 30, 3605–3635.
Oswald, A. J., Powdthavee, N., 2010. Daughters and Left-Wing Voting. The Review
of Economics and Statistics 92, 213–227.
– 24 –
Tate, G., Yang, L., 2015. Female leadership and gender equity: Evidence from plant
closure. Journal of Financial Economics 117, 77–97.
Warner, R. L., 1991. Does the Sex of Your Children Matter? Support for Feminism
among Women and Men in the United States and Canada. Journal of Marriage
and Family 53, 1051–1056.
Warner, R. L., Steel, B. S., 1999. Child Rearing as a Mechanism for Social Change:
The Relationship of Child Gender to Parents’ Commitment to Gender Equity.
Gender and Society 13, 503–517.
Washington, E. L., 2008. Female Socialization: How Daughters Affect Their Legisla-
tor Fathers. American Economic Review 98, 311–32.
– 25 –
Appendix: Definitions of Variables
Variable Definition
Firm-level variablesTreated firms An indicator variable that takes the value 1 for firms
with average employment between 35 and 50 in the pre-disclosure period (2003-2005) and 0 otherwise
Control firms An indicator variable that takes the value 1 for firmswith average employment between 35 and 50 in the pre-disclosure period (2003-2005) and 0 otherwise
Assets Measured in real USD. The source is KOB.Sales Measured in real USD.Source is KOB.Firm age Firm age based on the firm foundation date. The infor-
mation source is the business registry.Hierarchy We follow Caliendo, Monte, and Rossi-Hansberg (2015)
and Friedrich (2015) in constructing a measure on how hi-erarchical a firm is. The measure is based on the numberof different occupational layers represented by workers ina firm. We use workers’ occupations as reported in theDanish occupational code DISCO. The source is IDA
Employee-level variablesMale An indicator variable that takes the value 1 if the person
is male, and 0 otherwise. The source is the Danish CivilRegistration System.
Age Employee age. The source is the Danish Civil Registra-tion System. It is recoded into quartiles when using as aregression control.
Experience Employee’s number of years worked. It is recoded intoquartiles when using as a regression control.
No. of children The number of the employee’s living children. The sourceis the Danish Civil Registration System.
Wage Total annual wage of the employee. The informa-tion comes from the administrative-matched employer-employee dataset (IDA).
College degree An indicator variable that takes the value 1 if an em-ployee has completed a bachelor’s degree. The variableis constructed based on information from the official Dan-ish registry.
Promotion An indicator variable that takes the value 1 if the em-ployee got a promotion that year, and 0 otherwise. Thepromotion variable is constructed based on informationof employee position from IDA.
Separation An indicator variable that takes the value 1 if the em-ployee left the company that year, and 0 otherwise. Theseparation variable is constructed based on informationfrom IDA.
– 26 –
Table 1: Summary Statistics
This table reports summary statistics for the employee-level (Panel A) and firm level (Panel B) variablesfor all firms in our sample and for treated and control firms separately. Treated firms are those withaverage employment between 35 and 50 and controls are those with average employment between 20 and34, in the pre-law period (2003-2005). The variables are averaged over the pre-law years 2003-2005.The table reports unconditional means, medians and standard deviations. For the conversion fromDKK to USD we use the spot exchange rate at the year-end. Firm-level variables (except employmentand female shares) are winsorized at 1%.
Panel A - Employee-Level Characteristics
All Treated Control t-test
Observations Mean S.D. Mean S.D. Mean S.D. p-value
Wage (thous. $) 66,195 53.80 23.71 54.59 23.82 53.27 23.59 0.020
Hourly Wage ($) 66,188 33.92 15.09 34.41 15.38 33.54 14.82 0.013
Bonus (thous. $) 65,958 1.18 3.04 1.15 3.11 1.21 3.00 0.375
Age (years) 67,574 39.79 10.77 39.90 10.63 39.70 10.85 0.326
Male (%) 67,749 0.64 0.48 0.64 0.48 0.64 0.48 0.860
College degree (%) 66,158 0.25 0.43 0.25 0.44 0.24 0.43 0.213
Work Experience (years) 67,824 17.23 10.36 17.34 10.29 17.14 10.40 0.347
–27
–
Table 1: [Continued] Summary Statistics
Panel B1 - Firm-Level Characteristics
All Treated Control t-test
Observations Mean S.D. Mean S.D. Mean S.D. p-value
Assets (mil. $) 3,956 6.44 23.74 7.20 13.23 6.07 27.46 0.079
Sales (mil. $) 3,956 9.03 9.64 11.68 10.74 7.73 8.77 0.000
Sales/Employee (mil. $) 3,956 0.28 0.30 0.27 0.25 0.28 0.33 0.156
Employment 4,005 31.12 8.49 41.67 4.37 25.97 4.12 0.000
Female Share (%) 3,998 0.30 0.21 0.29 0.21 0.30 0.21 0.153
Profits (mil. $) 3,957 0.25 1.88 0.26 1.49 0.25 2.04 0.960
Profits/Employee (mil. $) 3,957 0.007 0.021 0.006 0.020 0.007 0.022 0.101
Wages (mil. $) 3,950 1.70 0.70 2.26 0.69 1.43 0.52 0.000
Wage/Employee (mil. $) 3,950 0.051 0.017 0.051 0.016 0.051 0.017 0.923
Pension & Soc. Sec. (mil. $) 3,950 0.135 0.082 0.179 0.091 0.114 0.068 0.000
Pension & Soc. Sec./Employee (mil. $) 3,950 0.004 0.002 0.004 0.002 0.004 0.002 0.819
Panel B2 - Pre-law Male Wage Premium
All Treated Control Difference
Male Wage Premium 0.240*** 0.257*** 0.226*** 0.031***
(No Controls) (0.0056) (0.0094) (0.0068) (0.0115)
Male Wage Premium 0.189*** 0.202*** 0.179*** 0.019***
(Full Specification Controls) (0.0041) (0.0063) (0.0054) (0.0072)
–28
–
Table 2: Univariate Test: Change in Compensation Policy Around the DisclosureLaw
This table reports the difference in average wage around the disclosure law for male andfemale employees. Column (1) pertain to employees of firms in the treated group and column(2) pertains to employees of control firms. Column (3) presents the difference betweenColumn (1) and Column (2) (difference-in-differences). The first row reports the differenceof average male wage between the post-law (2006-2008) and pre-law (2003-2005) periods forthe control (Column 1) and treated group (Column 2), and the difference between Column 1and Column 2 (Column 3). The second row similarly reports the first and second differencefor the average female wage. The third row reports the difference-in-difference-in-differencesresult, the difference between the change in the male wages and female wages around thedisclosure law, in treated versus control firms. Firms with average employment between 35and 50 in the pre-law period (2003-2005) are identified as treated, and firms with 20-34 areidentified as the control group. The wages are log-transformed. *** corresponds to statisticalsignificance at the 1% level. Standard errors are clustered at the firm level.
log Wage Treated Control Dif-in-Dif (DD)
(3-year avg after – 3-year avg before)
Male 0.0810*** 0.0967*** -0.0157***
(0.0034) (0.0029) (0.0044)
Female 0.1015*** 0.0981*** 0.0034
(0.0040) (0.0033) (0.0051)
Difference-in-Difference-in-Differences (DDD) -0.0190***
(0.0061)
– 29 –
Table 3: Gender Pay Gap Disclosure and Employee Wages
This table reports the effects of gender pay gap disclosure on employee wages. The dependent variableis employee annual wage. Treated firms are those with average employment between 35 and 50 andcontrols are those with average employment between 20 and 34, in the pre-law period (2003-2005).Post is 0 for years 2003-2005 and 1 for years 2006-2008. Person controls include employee experienceand age. All variables are defined in the Appendix. ***, **, and * correspond to statistical significanceat the 1%, 5%, and 10% levels, respectively. Standard errors are clustered at the firm level.
Male Female All Male Female All
Treated × Post -0.0167∗∗∗ 0.0028 0.0028 -0.0142∗∗∗ 0.0039 0.0040
(0.0039) (0.0045) (0.0044) (0.0035) (0.0042) (0.0042)
Male × Post -0.0022 -0.0030
(0.0034) (0.0033)
Treated × Post × Male -0.0195∗∗∗ -0.0185∗∗∗
(0.0052) (0.0050)
log(Sales) 0.0217∗∗∗ 0.0195∗∗∗ 0.0210∗∗∗
(0.0028) (0.0035) (0.0025)
Person-Firm FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Person Controls No No No Yes Yes Yes
Observations 145,852 79,532 225,384 145,262 79,027 224,289
R2 0.868 0.827 0.866 0.871 0.828 0.868
–30
–
Table 4: Gender Pay Gap Disclosure and Employee Wages: Treatment by Year
This table reports the Treated × Y ear effects of gender pay gapdisclosure on employee wages. The sample and variable definitionsfollow Table 3. Male × Y ear terms are estimated but omitted forbrevity. ***, **, and * correspond to statistical significance at the1%, 5%, and 10% levels, respectively. Standard errors are clusteredat the firm level.
Male Female All
Treated × Year2004 -0.0001 -0.0076 -0.0077
(0.0035) (0.0051) (0.0051)
Treated × Year2005 -0.0054 -0.0059 -0.0060
(0.0040) (0.0057) (0.0057)
Treated × Year2006 -0.0140∗∗∗ -0.0033 -0.0033
(0.0047) (0.0062) (0.0062)
Treated × Year2007 -0.0193∗∗∗ 0.0016 0.0017
(0.0053) (0.0066) (0.0066)
Treated × Year2008 -0.0185∗∗∗ -0.0004 -0.0004
(0.0059) (0.0072) (0.0071)
Male × Treated × Year2004 0.00748
(0.00602)
Male × Treated × Year2005 0.0005
(0.0066)
Male × Treated × Year2006 -0.0110
(0.0072)
Male × Treated × Year2007 -0.0213∗∗∗
(0.0078)
Male × Treated × Year2008 -0.0185∗∗
(0.0084)
log(Sales) 0.0218∗∗∗ 0.0196∗∗∗ 0.0210∗∗∗
(0.0028) (0.0035) (0.0025)
Person-Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Person Controls Yes Yes Yes
Observations 145,262 79,027 224,289
R2 0.871 0.828 0.868
– 31 –
Table 5: Gender Pay Gap Disclosure and Employee Wages: Placebo tests
This table reports a placebo estimation of gender pay gap disclosure on employee wages. In columns 1-2, the placebo treatment groupincludes firms with average employment of 20-35 and the placebo control group includes firms with average employment of 5-19 in thepre-treatment years 2003-2005. In columns 3-4, the ranges are 50-65 and 35-49 employees, respectively. In columns 5-6, the ranges are65-80 and 50-64 employees, respectively. The sample and variable definitions follow Table 3. ***, **, and * correspond to statisticalsignificance at the 1%, 5%, and 10% levels, respectively. Standard errors are clustered at the firm level.
20 Cutoff 50 Cutoff 65 Cutoff
Male Female All Male Female All Male Female All
Treatedp × Post 0.0039 0.0017 0.0017 -0.0009 -0.0018 -0.0017 0.0019 0.0011 0.0009
(0.0037) (0.0040) (0.0040) (0.0045) (0.0056) (0.0055) (0.0072) (0.0067) (0.0066)
Male × Post -0.0062 -0.0223∗∗∗ -0.0198∗∗∗
(0.0040) (0.0039) (0.0051)
Treatedp × Post × Male 0.0022 0.0011 0.0007
(0.0051) (0.0063) (0.0089)
log(Sales) 0.0210∗∗∗ 0.0215∗∗∗ 0.0212∗∗∗ 0.0204∗∗∗ 0.0149∗∗∗ 0.0182∗∗∗ 0.0405∗∗ 0.0151∗∗∗ 0.0291∗∗∗
(0.0026) (0.0030) (0.0022) (0.0035) (0.0038) (0.0031) (0.0174) (0.0048) (0.0112)
Person-Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Person Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 148,573 88,160 236,733 104,098 56,899 160,997 72,578 41,786 114,364
R2 0.865 0.827 0.863 0.875 0.822 0.871 0.862 0.815 0.859
–32
–
Table 6: Gender Pay Gap Disclosure and Employee Wages by Hierarchy
This table reports the effects of gender pay gap disclosure on employee wages, by employee position in the firm hierarchy. The sampleand variable definitions follow Table 3. ***, **, and * correspond to statistical significance at the 1%, 5%, and 10% levels, respectively.Standard errors are clustered at the firm level.
High-level Intermediate-level Lower-level
Male Female All Male Female All Male Female All
Treated × Post -0.0108 -0.0017 -0.0008 -0.0208∗∗∗ 0.0055 0.0057 -0.0106∗∗ 0.0029 0.0028
(0.0081) (0.0132) (0.0130) (0.0056) (0.0071) (0.0070) (0.0046) (0.0054) (0.0054)
Male × Post -0.0190∗ 0.0104∗ -0.0063
(0.0104) (0.0056) (0.0044)
Treated × Post × Male -0.0101 -0.0268∗∗∗ -0.0137∗∗
(0.0145) (0.0083) (0.0066)
log(Sales) 0.0228∗∗∗ 0.0190∗∗ 0.0220∗∗∗ 0.0220∗∗∗ 0.0178∗∗∗ 0.0206∗∗∗ 0.0209∗∗∗ 0.0210∗∗∗ 0.0209∗∗∗
(0.0054) (0.0092) (0.0055) (0.0040) (0.0059) (0.0037) (0.0044) (0.0043) (0.0034)
Person-Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Person Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 33,647 9,146 42,793 45,901 27,056 72,957 61,136 39,663 100,799
R2 0.829 0.805 0.829 0.849 0.799 0.849 0.856 0.810 0.847
–33
–
Table 7: Gender Pay Gap Disclosure and Employee Hiring
This table reports the effects of gender pay gap disclosure on the firm’s join-ing rate and leaving rate of employees. In Panel A, Joining Rate is de-fined as # female employees joining in year t
# total employees joining in year t . In Panel B, Leaving Rate is defined as# female employees leaving in year t# total employees leaving in year t . The sample is defined at the firm level. All variablesare defined in the Appendix. ***, **, and * correspond to statistical significance atthe 1%, 5%, and 10% levels, respectively. Standard errors are clustered at the firmlevel.
Panel A - Joining Rate
High-level Intermediate-level Lower-level
Treated × Post 0.0091 0.0088 0.0423∗∗ 0.0441∗∗ 0.0257 0.0248
(0.0248) (0.0251) (0.0216) (0.0217) (0.0192) (0.0192)
log(Sales) 0.0018 -0.0079 0.0201
(0.0160) (0.0145) (0.0137)
Firm FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Observations 3,221 3,208 5,391 5,373 7,046 7,035
R2 0.500 0.500 0.533 0.533 0.555 0.555
Panel B - Leaving Rate
High-level Intermediate-level Lower-level
Treated × Post 0.0216 0.0175 0.00765 0.00779 -0.0138 -0.0130
(0.0242) (0.0243) (0.0208) (0.0208) (0.0185) (0.0185)
log(Sales) 0.0149 -0.00406 0.0151
(0.0149) (0.0148) (0.0145)
Firm FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Observations 3,698 3,673 5,753 5,735 7,840 7,825
R2 0.465 0.467 0.516 0.517 0.564 0.564
– 34 –
Table 8: Gender Pay Gap Disclosure and Employee Promotion
This table reports the effects of gender pay gap disclosure on employee promotion likelihood.We consider an employee is promoted if he/she works at a higher hierarchy level that year,and 0 otherwise. Columns 1-3 show the results for employees who were in the intermediatehierarchy level in the previous year, and columns 4-6 show the results for those who werein the low hierarchy level in the previous year. The sample and variable definitions followTable 3. ***, **, and * correspond to statistical significance at the 1%, 5%, and 10% levels,respectively. Standard errors are clustered at the firm level.
Intermediate-level Low-level
Male Female All Male Female All
Treated × Post 0.0067 -0.0019 -0.0021 0.0019 0.0116∗∗ 0.0115∗∗
(0.0047) (0.0042) (0.0042) (0.0042) (0.0050) (0.0050)
Male × Post 0.0011 0.0010
(0.0034) (0.0032)
Treated × Post × Male 0.0087 -0.0097∗
(0.0060) (0.0052)
log(Sales) -0.0027 -0.0023 -0.0025 -0.0018 -0.0020 -0.0019
(0.0024) (0.0023) (0.0019) (0.0034) (0.0039) (0.0029)
Person-Firm FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Person Controls Yes Yes Yes Yes Yes Yes
Observations 35,166 19,907 55,073 52,382 33,398 85,780
R2 0.429 0.380 0.417 0.522 0.527 0.524
– 35 –
Table 9: Gender Pay Gap Disclosure and Firm Performance
This table reports the effects of gender pay gap disclosure on firm’s performance. In columns 1-2, the dependentvariable is the logarithm, of sales per employee; in columns 3-4, the dependent variable is the logarithm of wagesper employee; in columns 5-6, the dependent variable is the logarithm of Pension & Social Security expensesper employee; in columns 7-8, the dependent variable is profits per employee. The sample is defined at the firmlevel. All variables are defined in the Appendix. ***, **, and * correspond to statistical significance at the 1%,5%, and 10% levels, respectively. Standard errors are clustered at the firm level.
log(Sales/employees) log(Wage/employees) log(Pension &Soc.Sec./employees) Profits/employees
Treated × Post -0.0250∗∗ -0.0246∗∗ -0.0282∗∗∗ -0.0281∗∗∗ -0.0018 -0.0034 6.22 6.11
(0.0120) (0.0112) (0.0056) (0.0053) (0.0207) (0.0207) (3.79) (3.73)
log(Sales) 0.4050∗∗∗ -0.1040∗∗∗ 0.1010∗∗∗ 26.26∗∗∗
(0.0224) (0.0113) (0.0186) (5.19)
Firm FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Observations 22,414 22,391 22,429 22,391 22,374 22,351 21,602 21,564
R2 0.845 0.879 0.849 0.863 0.544 0.547 0.621 0.625
–36
–
Table 10: Mechanisms: Managerial Preferences
This table reports the effects of gender pay gap disclosure and its interactionwith whether managers have pro-women preferences on employee wages. Weexclude female managers and consider (up to) five male managers in the pre-treatment (2003-2005) period. We define a variable to be 1 if a male managerhas more daughters than sons, 0.5 if they have as many daughters as sons,and 0 otherwise. Female Child is an indicator that takes a value of 1 if thefirm average exceeds the sample median, and 0 otherwise. The sample andvariable definitions follow Table 3. ***, **, and * correspond to statisticalsignificance at the 1%, 5%, and 10% levels, respectively. Standard errors areclustered at the firm level.
Male Female All
Tested×Post -0.0184*** -0.0059 -0.0057
(0.0050) (0.0058) (0.0057)
Post×Female Child 0.0071 -0.0025 -0.0026
(0.0054) (0.0058) (0.0057)
Treated×Post×Female Child -0.0085 0.0158* 0.0157*
(0.0080) (0.0088) (0.0087)
Male×Post 0.0058
(0.0046)
Treated×Post×Male -0.0131*
(0.0068)
Post×Male×Female Child 0.0097
(0.0073)
Treated×Post×Male×Female Child -0.0243**
(0.0109)
Log(Sales) 0.0211*** 0.0197*** 0.0206***
(0.0030) (0.0035) (0.0026)
Person-Frim FE Yes Yes Yes
Year FE Yes Yes Yes
Person Controls Yes Yes Yes
Observations 122,266 74,516 196,782
R2 0.851 0.815 0.848
– 37 –
Table 11: Mechanisms: Industry Gender Pay Gap
This table reports the effects of gender pay gap disclosure and its interac-tion with pre-treatment industry gender pay gap on employee wages. We de-fine Ind. Gender Gap at the industry-occupation level by computing the me-dian log difference in wages by gender at the industry-occupation-year leveland averaging over the pre-treatment period (2003-2005). Ind. Gender Gap,Treated×Ind.Gender Gap, Male×Ind.Gender Gap are estimated but not reportedfor brevity. The sample and variable definitions follow Table 3. ***, **, and *correspond to statistical significance at the 1%, 5%, and 10% levels, respectively.Standard errors are clustered at the firm level.
Male Female All
Treated×Post -0.0146∗∗∗ 0.0055 0.0056
(0.0037) (0.0043) (0.0043)
Post×Ind. Gender Gap -0.0017 -0.0012 -0.0016
(0.0040) (0.0046) (0.0045)
Treated×Post×Ind. Gender Gap -0.0032 0.0125∗∗ 0.0126∗∗
(0.0059) (0.0060) (0.0060)
Male×Post -0.0020
(0.0034)
Treated×Post×Male -0.0205∗∗∗
(0.0051)
Treated×Male×Ind. Gender Gap -0.0100
(0.0165)
Post×Male×Ind. Gender Gap -0.0003
(0.0058)
Treated×Post×Male×Ind. Gender Gap -0.0159∗∗
(0.0079)
log(Sales) 0.0217∗∗∗ 0.0193∗∗∗ 0.0208∗∗∗
(0.0028) (0.0035) (0.0026)
Person-Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Person Controls Yes Yes Yes
Observations 138,576 77,609 216,185
R2 0.871 0.828 0.868
– 38 –
Online Appendix to
"Do firms respond to gender pay gap disclosure?"
Morten Bennedsen, Elena Simintzi, Margarita Tsoutsoura, and Daniel Wolfenzon
– 39 –
Table IA1: Robustness: Sample restriction to control for employee compositionchanges
This table repeats our baseline specification including in the sam-ple only employees that were working for the firm at least one fullyear before the law and one full year after. Years where employeesjoined or left are not counted as full years. The variable definitionsfollow Table 3. ***, **, and * correspond to statistical significanceat the 1%, 5%, and 10% levels, respectively. Standard errors areclustered at the firm level.
Male Female All
Treated × Post -0.0097*** 0.0058* 0.0059*
(0.0028) (0.0033) (0.0033)
Male × Post -0.0002
(0.0026)
Treated × Post × Male -0.0159***
(0.0039)
Log(Sales) 0.0183*** 0.0144*** 0.0170***
(0.0023) (0.0029) (0.0021)
Person-Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Person Controls No No No
Observations 94,118 49,451 143,569
R2 0.933 0.894 0.929
– 40 –
Table IA2: Robustness: Firm-Year Fixed Effects
This table repeats our baseline specification and the Treated×Y eareffects of gender pay gap additionally controlling for firm-year fixedeffects. The sample restrictions and variable definitions follow Ta-ble 3. ***, **, and * correspond to statistical significance at the1%, 5%, and 10% levels, respectively. Standard errors are clusteredat the firm level.
Baseline Treatment by year
All All
Male × Post -0.0044
(0.0035)
Treated × Post × Male -0.0143∗∗∗
(0.0051)
Male × Treated × Year2004 0.0048
(0.0066)
Male × Treated × Year2005 -0.0072
(0.0073)
Male × Treated × Year2006 -0.0116
(0.0077)
Male × Treated × Year2007 -0.0202∗∗
(0.0083)
Male × Treated × Year2008 -0.0200∗∗
(0.0088)
Person-Firm FE Yes Yes
Year FE Yes Yes
Firm-Year FE Yes Yes
Person Controls Yes Yes
Observations 222,529 222,529
R2 0.885 0.885
– 41 –
Table IA3: Robustness: Employee Hourly Wage
This table reports the effects of gender pay gap disclosure on employee hourly wages. The sampleand variable definitions follow Table 3. ***, **, and * correspond to statistical significance at the1%, 5%, and 10% levels, respectively. Standard errors are clustered at the firm level.
Male Female All Male Female All
Treated × Post -0.0130∗∗∗ 0.0001 0.0001 -0.0116∗∗∗ 0.0008 0.0008
(0.0031) (0.0034) (0.0034) (0.0030) (0.0033) (0.0033)
Male × Post 0.0078∗∗∗ 0.0074∗∗∗
(0.0026) (0.0026)
Treated × Post × Male -0.0130∗∗∗ -0.0125∗∗∗
(0.0039) (0.0039)
log(Sales) 0.0131∗∗∗ 0.0139∗∗∗ 0.0135∗∗∗
(0.0026) (0.0035) (0.0026)
Person-Firm FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Person Controls No No No Yes Yes Yes
Observations 153,062 83,895 236,957 152,460 83,372 235,832
R2 0.906 0.884 0.907 0.907 0.886 0.908
– 42 –
Table IA4: Robustness: Employee Wages and Bonus Payments
This table reports the effects of gender pay gap disclosure on employee wages and bonus payments.The sample and variable definitions follow Table 3. ***, **, and * correspond to statistical significanceat the 1%, 5%, and 10% levels, respectively. Standard errors are clustered at the firm level.
Male Female All Male Female All
Treated × Post -0.0146∗∗∗ 0.0030 0.0030 -0.0120∗∗∗ 0.0041 0.0043
(0.0040) (0.0046) (0.0046) (0.0037) (0.0044) (0.0044)
Male × Post -0.0020 -0.0028
(0.0036) (0.0034)
Treated × Post × Male -0.0176∗∗∗ -0.0166∗∗∗
(0.0054) (0.0051)
log(Sales) 0.0234∗∗∗ 0.0206∗∗∗ 0.0224∗∗∗
(0.0030) (0.0037) (0.0027)
Person-Firm FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Person Controls No No No Yes Yes Yes
Observations 144,811 79,001 223,812 144,235 78,510 222,745
R2 0.866 0.828 0.865 0.869 0.829 0.867
– 43 –
Table IA5: Top 5 Earners Female First Child Ratios
This table repeats Table 10, except focusing on managers’ first child to constructFemale Chile variable.***, **, and * correspond to statistical significance at the1%, 5%, and 10% levels, respectively. Standard errors are clustered at the firm level.
Male Female All
Tested×Post -0.0221*** -0.0084 -0.0080
(0.0051) (0.0059) (0.0059)
Post×Female Child Ratio 0.0011 -0.0047 -0.0047
(0.0054) (0.0057) (0.0057)
Treated×Post×Female Child Ratio -0.0001 0.0203** 0.0199**
(0.0079) (0.0087) (0.0087)
Post×Male 0.0074
(0.0049)
Treated×Post×Male -0.0146**
(0.0071)
Post×Male×Female Child Ratio 0.0059
(0.0073)
Treated×Post×Male×Female Child Ratio -0.0197*
(0.0107)
Log(Sales) 0.0211*** 0.0197*** 0.0206***
(0.0030) (0.0035) (0.0026)
Person-Firm FE Yes Yes Yes
Year FE Yes Yes Yes
Person Controls Yes Yes Yes
Observations 122,232 74,528 196,760
R2 0.851 0.816 0.848
– 44 –
Table IA6: Gender Pay Gap Disclosure and Firm Performance: Dynamics
This table is the by-year version of Table 11: Gender Pay Gap Disclosure and Firm Performance. This table reportsthe effects of gender pay gap disclosure on firm’s performance. In columns 1-2, the dependent variable is the logarithm,of sales per employee; in columns 3-4, the dependent variable is the logarithm of wages per employee; in columns5-6, the dependent variable is the logarithm of Pension & Social Security expenses per employee; in columns 7-8, thedependent variable is profits per employee. The sample is defined at the firm level. All variables are defined in theAppendix. ***, **, and * correspond to statistical significance at the 1%, 5%, and 10% levels, respectively. Standarderrors are clustered at the firm level.
log(Sales/employees) log(Wage/employees) log(Pension &Soc.Sec./employees) Profits/employees
Treated × Year2004 -0.0008 -0.0039 -0.0023 -0.0014 0.0206 0.0210 -0.0725 -0.433
(0.0137) (0.0108) (0.0041) (0.0042) (0.0293) (0.0291) (3.661) (3.669)
Treated × Year2005 0.0005 -0.0102 -0.0076 -0.0058 0.0314 0.0298 -2.886 -3.494
(0.0159) (0.0135) (0.0061) (0.0060) (0.0323) (0.0324) (4.504) (4.471)
Treated × Year2006 -0.0045 -0.0113 -0.0193*** -0.0181*** 0.0441 0.0414 6.197 5.453
(0.0176) (0.0152) (0.0069) (0.0067) (0.0341) (0.0339) (5.201) (5.200)
Treated × Year2007 -0.0395** -0.0419** -0.0379*** -0.0376*** 0.0079 0.0064 4.477 4.031
(0.0183) (0.0168) (0.0081) (0.0076) (0.0332) (0.0331) (5.500) (5.470)
Treated × Year2008 -0.0327* -0.0364** -0.0386*** -0.0371*** -0.0071 -0.0088 4.837 4.735
(0.0192) (0.0183) (0.0091) (0.0085) (0.0364) (0.0364) (6.202) (6.071)
log(Sales) 0.4050∗∗∗ -0.1040∗∗∗ 0.1010∗∗∗ 26.27∗∗∗
(0.0224) (0.0113) (0.0187) (5.189)
Firm FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Observations 22,414 22,391 22,429 22,391 22,374 22,351 21,602 21,564
R2 0.845 0.879 0.849 0.863 0.544 0.547 0.621 0.625
–45
–